Hi,
on 01.11.2023 Nvidia released the 545 version of their proprietary drivers. My freshly updated Rocky 9 running fine with the 535 both in Wayland and Xorg ceased to function (in Xorg). Wayland is sluggish and uses a lot of my card memory (GTX 1070) triggering a “GPU Memory Full” on Davinci Resolve 18.6. Downgrading to 5.3.5 seems to solve the issue but i could not pinpoint the problem since the session seem to start, but cannot find settings for display and then hangs. A switch to a command prompt (CTRL+ALT+F2) and back, gets me back to login prompt but i cannot select anymore the server i want to start (the cog on the lower right corner of the screen is missing too), Only a reboot solves the problem. Could not find anything on how to fix/diagnose the issue since everything seems to be normal in the logs, i am running on a desktop machine not a Laptop with Dual Nvidia/Intel card so the only solution for now is to stay with 535 until better drivers or a fix comes up.
Use the drivers from RPMFusion. Howto/NVIDIA - RPM Fusion
I hadn’t heard of Negativo17 until now, and looking at his website, he only mentions EL7, not EL9.
Hi,
Same issue with the drivers on RPMFusion.
Negativo had them packed nicely when RPMFusion and others, including the ones from NVIDIA were having issues. and they do have epel9 see here Index of /repos/nvidia
thanks for the suggestion though.
Regards,
JD
I did find it when EL9 was new. It had something a bit before RPMFusion. I would have preferred ELRepo, but they chose not to build NVidia for el9. (I do use NVidia’s own repo for current cards and RPMFusion for legacy as it is the only to build legacy series that I know and not even all of them.)
Tristan Theroux wrote yesterday (in AlmaLinux chat):
Reporting this here since Nvidia is not so transparent with the Linux community.
Support for Nvidia Enterprise confirmed to me that in the latest releases of Nvidia drivers for linux, there is a known issue we’re the kernel, does not get rebuild correctly even with DKMS installed. It will tell you that there’s a Driver/Library mismatch.
This is due to the fact that the .run driver does not update the kernel totally…
The fix is now to input a
dracut -f, after your installation right before the reboot.
That sounds like a different issue … (?)
Thanks so much for the info. Checked on the “www” and it looks like the issue was reported also for 535 and 512/513. It doesn’t look to be this particular issue since i cannot find the “failed to load” message in boot.log. Since this the PC i am having issues with is production machine, i am trying to stay away from constantly updating/downgrading it to test different scenarios, given the fact that i kind’a need it up and running for editing/VFX work. What i will try to do later is to switch repos from negativo to RPMfusion, attempt again to install nvidia drivers from there and do the dracut, just in case there might be some hidden things i missed.
thanks so much.
JD
Quick update,
i can confirm that following the tutorial here : How to Install NVIDIA Drivers on Rocky Linux EL9 or EL8 - LinuxCapable i can install latest nvidia 545 drivers. i also issued the infamous “dracut -f” command before rebooting and now i have everything running again in Xorg!
Unfortunately i don’t have time to dive into the what’s ans why’s the negativo drivers failed but for now i switched to official and i’m happy. Thanks so much for all your help.
Cheers,
JD